• Keine Ergebnisse gefunden

Essays on the Econometric Analysis of Energy Markets and Climate Change

N/A
N/A
Protected

Academic year: 2021

Aktie "Essays on the Econometric Analysis of Energy Markets and Climate Change"

Copied!
235
0
0

Wird geladen.... (Jetzt Volltext ansehen)

Volltext

(1)

D

OCTORAL

T

HESIS

Essays on the Econometric Analysis of Energy

Markets and Climate Change

Author:

Christoph FUNK

Supervisor:

Prof. Dr. Peter WINKER

Supervisor:

Prof. Dr. Peter TILLMANN

A thesis submitted in fulfillment of the requirements for the degree of Doctor rerum politicarum

in the

Department of Statistics and Econometrics Faculty of Economics and Business Studies

Submission: July 29, 2020 Defended on: January 18, 2021

(2)
(3)

Statement of Originality

This thesis is being submitted to Macquarie University and Justus Liebig University Giessen in accordance with the Cotutelle agreement dated March 12, 2018.

To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.

Christoph FUNK

(4)
(5)

JUSTUSLIEBIGUNIVERSITYGIESSEN

Abstract

Department of Statistics and Econometrics Faculty of Economics and Business Studies

Doctoral Thesis

Essays on the Econometric Analysis of Energy Markets and Climate Change

by Christoph FUNK

This PhD thesis uses econometric techniques to provide new insights into energy markets and climate change. The developed econometric models are applied in six research articles that focus on energy markets, financial economics, climate vulnerability and adaptation strategies, and the impacts of climate change on coastal regions. The first part of the thesis is dedicated to the econometric analysis of the oil market. Crude oil and its derivatives are the world’s most commonly traded commodities. Although the energy landscape is in a state of transi-tion, crude oil has been and will likely remain important in the context of global energy con-sumption. By focusing on econometric and time series regression models, Chapter 2 examines whether it is possible to generate reliable long-term monthly forecasts of the real price of crude oil and how these can be improved using forecast combinations. Chapter 3 investigates the North American oil industry using firm-level data. The relationship between oil prices, oil market volatility, and the cost of debt for oil firms is analyzed empirically using a distributed lag model and a panel data “within-between” approach. The second part of the thesis focuses on climate change vulnerability and adaptation strategies in the context of smallholder farmers in the Indian watersheds. The global climate has changed relative to the pre-industrial period. Moreover, this change already affects ecosystems and livelihoods worldwide. This has severe implications for food production and security, especially for the rain-fed agricultural systems that dominate much of tropical agriculture and are extremely vulnerable to projected climate change. Consequently, Chapters 4 and 5 focus on two research questions related to watershed development programs and factors that affect the adaptation strategies of smallholder farmers in India. Chapter 4 provides a theoretical framework for developing a climate vulnerability index. Chapter 5 is aimed at investigating how smallholder farmers perceive climate change and how they adapt their behavior in response to perceived changes in the climate. Using a binary logistic model, this thesis quantifies the impact of various explanatory variables that affect households’ choices of adaptation strategies. The final part of this thesis investigates the impacts of climate change on coastal regions. Anthropogenic climate change has caused global mean sea levels to rise substantially over the last century. As a result, local and regional sea level variations and the occurrence of sea level extremes increase climate change related risks for coastal regions. This amplifies the need to understand how waves and extremes change over the long term to assess the impacts of climate change. Thus, Chapter 6 provides a lit-erature review and guide for policy makers on how to evaluate coastal adaptation projects. Moreover, Chapter 7 empirically investigates long-term trends in and extreme values of wave power and height on a local scale along Australia’s east and southeast coasts. A comparison of the distribution of wave power employing the Jensen-Shannon divergence and Laplacian embedding is used as a potential tool for investigating temporal and spatial variations of the wave climate between and within different locations. This provides new insights on Australia’s wave climate on a local and regional scale in the context of climate change.

(6)
(7)

Acknowledgements

This cumulative thesis was written as part of a Cotutelle PhD program between Justus Liebig University Giessen in Germany and Macquarie University in Australia. I feel quite privileged to have had the opportunity to work and conduct research at these two excellent research in-stitutions. I would like to thank Macquarie University for providing me with a Macquarie University Research Excellence Scholarship (iMQRES). In addition, this thesis would not have been possible without the support of many others whom I would like to thank.

First, I would like to express my sincere gratitude to my supervisors and mentors, Peter Winker, Stefan Trück, and Peter Tillmann, for their continuous support over the last years. I am very thankful to Peter Winker for providing me the opportunity to write my thesis at his chair and giving me the freedom to explore my research topics. Your commitment and encouragement made it possible to continue my research in Australia. Thank you for enabling and motivating me to take this great chance. I also would like to thank Stefan Trück for his encouragements, patience, and engagement during my time in Australia. You have been a great source of inspiration and help, not only in writing this thesis, but especially by being supportive during my low times. In addition, I am grateful that Peter Tillmann was willing to serve as my second supervisor in Germany. Thank you for your support throughout my time at Giessen and your engagement, which enabled me to conduct my research in Australia. I would also like to thank Jürgen Meckl, especially for giving me the opportunity to study in Milwaukee and for his support throughout my PhD at Giessen.

This PhD thesis also includes joint papers written with the following co-authors: Lutz Breuer, Peter Winker, Stefan Trück, Thomas Aenis, Archana Raghavan Sathyan, Chi Truong, Supriya Mathew, Johannes Lips, and Karol Kempa. I would like to thank them all for their constructive and rewarding cooperation.

I would like to thank all my colleagues at Giessen and Macquarie University for their sup-port. Special thanks go to my friends Johhannes Lips, Jana Brandt, and Daniel Grabowski for their valuable input and especially for creating a fun, engaging work environment. I am thank-ful to Archana Raghavan Sathyan for her cooperation, friendship, and support over the course of this PhD. Moreover, I would like to thank Carmen Hersener for always having an open ear for my problems during our morning coffee breaks and for her countless administrative support.

Last but surely not the least, I would like to thank all my friends, my family, and my fiancée for their love, patience, and support throughout this journey. Without you, this thesis would not have been possible. Thank you!

(8)
(9)

Contents

Statement of Originality iii

Abstract v

Acknowledgements vii

List of Figures xiii

List of Tables xv

List of Abbreviations xvii

I Introduction 1

1 Introduction 3

1.1 Developments in the US Oil Market . . . 5

1.2 Climate Change: Vulnerability and Adaptation Strategies . . . 9

1.3 Climate Change: Coastal Adaptation and Wave Power . . . 11

II The Oil Market 13 2 Forecasting the Real Price of Oil – Time-Variation and Forecast Combination 15 Abstract . . . 16

2.1 Introduction . . . 17

2.2 Selected Forecasting Models . . . 18

2.2.1 No-change Benchmark . . . 18

2.2.2 AR and ARMA . . . 18

2.2.3 Futures . . . 18

2.2.4 Forecasts based on Crude Oil Inventories . . . 19

2.2.5 VAR Model of the Global Oil Market . . . 19

2.2.6 Non-Oil Industrial Raw Materials . . . 20

2.2.7 Price Movements with Oil Sensitive Stocks . . . 20

2.2.8 Other Forecasting Methods . . . 21

2.3 Data Sources and Preparation . . . 21

2.4 Individual Model Performance . . . 23

2.4.1 Forecasting Results . . . 23

2.4.2 Individual Model Performance Over Time . . . 25

2.5 Forecast Combination Approaches . . . 28

2.6 Uncertainty Analysis . . . 30

2.7 Conclusion . . . 36

B Appendix to Chapter 2 . . . 37

(10)

B.2 Robustness Check . . . 39

B.3 Alternative Benchmarks . . . 43

3 Oil Price and Cost of Debt: Evidence from Loans and Corporate Bonds 45 Abstract . . . 46

3.1 Introduction . . . 47

3.2 Debt Financing and the Oil Industry . . . 48

3.2.1 Determinants of Sources and Costs of Debt . . . 48

3.2.2 Specifics of the Oil Industry . . . 49

3.3 Data Set and Variables . . . 51

3.3.1 Oil Firms . . . 51

3.3.2 Bank Loans and Bonds . . . 52

3.3.3 Oil Price . . . 54

3.3.4 Macroeconomic Environment . . . 54

3.4 Empirical Analysis . . . 54

3.4.1 Exploratory Data Analysis . . . 54

3.4.2 Estimation Approach . . . 58

3.4.2.1 Distributed Lag Model for Credit Spreads at Issuance . . . 58

3.4.2.2 Panel Data Methods for Continuous Credit Spreads . . . 59

3.4.3 Estimation Results . . . 60

3.4.3.1 Determinants of Credit Spreads at Issuance . . . 60

3.4.3.2 Determinants of Credit Spreads on the Secondary Market . . . . 62

3.5 Conclusion . . . 64

C Appendix to Chapter 3 . . . 66

C.1 Classification of SIC and NAICS Codes . . . 66

C.2 Definition of Variables . . . 67

C.3 Exploratory Data Analysis . . . 68

III Climate Change: Vulnerability and Adaptation Strategies 73 4 Climate Vulnerability in Rainfed Farming: Analysis from Indian Watersheds 75 Abstract . . . 76

4.1 Introduction . . . 77

4.2 Materials and Methods . . . 79

4.2.1 Climate Vulnerability Assessment Concepts . . . 79

4.2.2 Climate Vulnerability Index . . . 79

4.2.3 Study Area . . . 83

4.2.4 Watershed Development Programs . . . 84

4.2.5 House Hold Surveys . . . 84

4.3 Results and Discussion . . . 85

4.3.1 Dimensions of Vulnerability and the Climate Vulnerability Index . . . 85

4.3.2 Major Components and Indicators . . . 86

4.3.2.1 Adaptive Capacity . . . 86

4.3.2.2 Sensitivity . . . 88

4.3.2.3 Exposure . . . 89

4.3.2.4 Comparison of Major Components . . . 90

4.4 Conclusion . . . 91

(11)

5 Changing Climate - Changing Livelihood: Smallholder’s Perceptions and Adaptation

Strategies 101

Abstract . . . 102

5.1 Introduction . . . 103

5.2 Methodology . . . 105

5.2.1 Description of the study area . . . 105

5.2.2 Development of the Database and Empirical Model . . . 106

5.2.3 Choice of Explanatory Variables . . . 108

5.3 Results and Discussion . . . 111

5.3.1 Perception of and Adaptation Strategies to Climate Change . . . 111

5.3.2 Model Significance and Goodness of Fit . . . 114

5.3.2.1 Factors Affecting the Choice of Adaptation Strategies . . . 115

5.3.2.2 Age of household head . . . 115

5.3.2.3 Education of household head . . . 115

5.3.2.4 Gender of household head . . . 115

5.3.2.5 Household size . . . 117

5.3.2.6 Various aspects of wealth . . . 117

5.3.2.7 Farm size in hectares . . . 118

5.3.2.8 Well ownership . . . 118

5.3.2.9 Extension services . . . 118

5.3.2.10 Climate-change awareness . . . 118

5.4 Conclusion and Policy Implications . . . 119

E Appendix to Chapter 5 . . . 120

IV Climate Change: Coastal Adaptation and Wave Power 125 6 Project Valuation for Coastal Adaptation – A Literature Review 127 Abstract . . . 128

6.1 Introduction . . . 129

6.2 Overview of Valuation Methods . . . 130

6.3 Review of Relevant Concepts and Methods . . . 132

6.3.1 The Discount Rate . . . 132

6.3.2 Estimation of Non-Market Values . . . 133

6.3.2.1 Revealed Preference Methods . . . 134

6.3.2.2 Stated Preference Methods . . . 135

6.3.3 Benefit Transfer Methods . . . 136

6.3.4 Multi-Criteria Analysis . . . 136

6.3.4.1 General Framework and Applications of MCA . . . 137

6.3.4.2 Criteria Weight Selection and Elicitation . . . 138

6.4 Review of Climate Adaptation Studies in Coastal Areas . . . 139

6.4.1 CBA Studies . . . 139

6.4.2 Timing Studies . . . 142

6.5 Conclusion . . . 145

7 Long-Term Trends in the Australian Wave Climate 147 Abstract . . . 148

7.1 Introduction . . . 149

7.2 Data and Descriptive Statistics . . . 151

7.2.1 Characterization of Parameters . . . 153

(12)

7.3 Laplacian Embedding . . . 156

7.4 Long-Term Trends in Wave Power . . . 159

7.5 Decadal Change . . . 163 7.6 Conclusion . . . 166 G Appendix to Chapter 7 . . . 168 V Conclusion 179 7 Conclusion 181 Bibliography 187 Declaration of Authorship 215

(13)

List of Figures

1.1 Structure of the thesis . . . 4

1.2 Primary energy consumption by energy source . . . 5

1.3 Share of primary energy consumption by energy source . . . 6

2.1 Time-Variation of recursive MSPE ratio with a 1-, 12-, 18- and 24- month fore-casting horizon . . . 27

2.2 Interval forecasts based on quantiles with a 12 month forecasting horizon . . . . 34

2.3 Interval forecasts based on a quantile regression with a 12 month forecasting horizon . . . 35

2.4 Spread between Brent and WTI . . . 39

3.1 Development of WTI crude oil spot price. . . 50

3.2 Development of aggregate assets and debt in each part of the value chain and the resulting debt-to-asset ratios . . . 55

3.3 Boxplots of main firm characteristics . . . 56

3.4 Number of bonds and loans issued per year for all firms in the sample . . . 56

3.5 Loan and bond credit spreads at issuance . . . 57

3.6 Development of the average and median bond credit spread on the secondary market . . . 57

3.7 Number of loans issued per industry classification per year . . . 68

3.8 Number of bonds issued per industry classification per year . . . 68

3.9 Median bond credit spread at issuance bonds per industry classification per year 69 3.10 Average loan credit spread at issuance and average maturity of loan facilities per industry classification per year . . . 69

3.11 Average volume of loans issued per industry classification per year . . . 70

3.12 Average volume of bonds issued per industry classification per year . . . 70

3.13 Average credit spread of bonds traded on the secondary market per industry classification per year . . . 71

4.1 The composition of the CVIRFT . . . 80

4.2 Location of the study area in Kerala, India . . . 84

4.3 Dimensions of Vulnerability and the resulting CVIRFTof the three watersheds . . 86

4.4 Comparison of average values of the ten major components for the three water-sheds . . . 86

4.5 Correlation analysis of the 59 variables . . . 98

5.1 Households’ perceptions of climate change . . . 112

5.2 Seasonal and annual temperature and rainfall . . . 113

5.3 Households’ adaptation strategies . . . 114

5.4 ROC curves for out-of-sample predictions . . . 116

5.5 Coefficient plot for the regression results of table A.1. . . 121

5.6 Correlation table for variables of Table 5.1 . . . 122

(14)

6.2 Valuation techniques for use and non-use values . . . 135

7.1 Locations of wave measuring buoys . . . 151

7.2 WP exceedance . . . 155

7.3 Laplacian embedding of all stations . . . 157

7.4 Laplacian embedding of all Stations per year . . . 158

7.5 Laplacian embedding of all stations as averages per year . . . 158

7.6 Difference in median decadal WP . . . 164

7.7 Difference in 90% quantiles decadal WP . . . 165

7.8 Difference in 99% quantiles of decadal WP . . . 165

7.9 Gaps in daily data for buoys in Queensland . . . 168

7.10 Gaps in daily data for buoys in NSW, South Australia and Tasmania . . . 169

7.11 Scatterplots between Hs and Tp for Buoys in Queensland . . . 171

7.12 Scatterplots between Hs and Tp for Buoys in NSW, South Australia and Tasmania 172 7.13 Rainbow plots of WP for Queensland . . . 173

7.14 Rainbow plots of WP for NSW, South Australia and Tasmania . . . 174

7.15 Trend in annual WP of buoys in Queensland . . . 175

(15)

List of Tables

2.1 Description of data . . . 22

2.2 Recursive MSPE ratios relative to the no-change benchmark . . . 24

2.3 Recursive success ratios relative to the no-change benchmark . . . 25

2.4 Forecast efficiency . . . 26

2.5 Recursive MSPE ratios relative to the no-change benchmark for an all model and a 4 model combination . . . 29

2.6 Recursive success ratios relative to the no-change benchmark for an all model and a 4 model combination . . . 30

2.7 Leave one out sensitivity analysis of the 4 model combination . . . 31

2.8 Failure rate based on quantiles of previous out-of-sample forecast errors . . . 32

2.9 Failure rate based on a quantile regression approach using previous out-of-sample forecast errors . . . 33

2.10 Recursive MSPE ratios relative to the random walk benchmark . . . 37

2.11 Recursive success ratios relative to the random walk benchmark . . . 38

2.12 Robustness check - Recursive MSPE ratios relative to the no-change benchmark for WTI . . . 40

2.13 Recursive success ratios relative to the no-change benchmark . . . 41

2.14 Robustness check - Leave one out sensitivity analysis of the 4 model combination relative to the WTI benchmark . . . 42

2.15 Recursive MSPE ratios relative to a benchmark computed using the average growth rate of the real price of Brent oil . . . 43

2.16 Recursive MSPE ratios relative to an AR Benchmark . . . 44

3.1 Summary statistics - Full sample for all firms . . . 53

3.2 Determinants of the loan credit spread at issuance. . . 61

3.3 Determinants of the bond credit spread at issuance. . . 62

3.4 Within-between effects estimation of the determinants of the bond credit spread on the secondary market for the full sample. . . 63

3.5 Information on the estimated within-between model. . . 64

3.6 Recursive success ratios relative to the no-change benchmark . . . 66

3.7 Definition of variables used in the empirical analysis. . . 67

4.1 Dimensions of Vulnerability, their ten major components and the 59 indicators involved for estimation of the CVIRFT . . . 81

4.2 Socio-economic and physical characteristics of the selected watersheds and the implemented WDPs . . . 85

4.3 Indexed values for the major components of adaptive capacity . . . 87

4.4 Indexed values for the major components of sensitivity . . . 88

4.5 Indexed values for the major components of exposure . . . 90

4.6 List of indicators, their explanation and sources . . . 92

4.7 Indicators of major components with its actual (A) and hypothesised (H) values for the watersheds. . . 95

(16)

5.1 Description of the variables . . . 110

5.2 The models’ significance and goodness of fit . . . 115

5.3 Regression results – Odds ratios . . . 117

5.4 Regression results – Odds ratios (full table) . . . 120

5.5 Regression results – Marginal effects reported as the data’s sample mean . . . 123

5.6 Regression results – Marginal effects ereported as the mean of the marginal ef-fects across all observations . . . 124

7.1 Descriptive statistics of the available data set . . . 152

7.2 Descriptive statistics for WP . . . 155

7.3 Trend analysis for Queensland of annual WP . . . 161

7.4 Trend Analysis for NSW, South Australia and Tasmania of annual WP . . . 162

7.5 Tukey difference-in-mean test . . . 163

7.6 Descriptive statistics for the Hs . . . 170

7.7 Descriptive statistics for Tp . . . 170

7.8 Trend analysis for Queensland of annual Hs . . . 177

(17)

List of Abbreviations

2D Two-dimensional

A Agriculture

AR Autoregressive

ARMA Autoregressive Moving Average

AUC Area Under the Curve

Brent Brent North Sea Crude Oil

Btu British Thermal Unit

CBA Cost Benefit Analysis

CEA Cost Effective Analysis

CEAS Coastal Erosion Adaptation Strategies

CPI Consumer Price Index

CRB Commodity Research Bureau

CUA Cost Utility Analysis

CUSIP Committee On Uniform Security Identification Proce-dures

CV Climate Variability

CVIRFT Climate Vulnerability Index for Rainfed Tropics EIA U.S. Energy Information Administration

F Food

FE Fixed Effects

FINRA Financial Industry Regulatory Authority

FRASER Federal Reserve Archival System for Economic Research

FRED Federal Reserve Economic Data

GMSL Global Mean Sea Level

H Health

Hs Significant Wave Height

JONSWAP Joint North Sea Wave Observation Project

JSD Jensen–Shannon Divergence

LG Local-Self Government

LIBOR London Interbank Offering Rate

LS Livelihood Strategies

MCA Multi-criteria Analysis

(18)

NAICS North American Industry Classification System

ND Natural Disaster Impact

NGO Non-Governmental Organisation

NPV Net Present Value

NSW New South Wales

NYSE New York Stock Exchange

OECD Organisation for Economic Co-operation and Develop-ment

OPEC Organization of the Petroleum Exporting Countries

OTC Over-the-counter

QE Quantitative Easing

RAC U.S. Refiners’ Acquisition Cost

RE Random Effects

ROA Real Options Analysis

ROC Receiver Operating Characteristic

SD Socio-Demographic Profile

SDGs Sustainable Development Goals

SE Socio-Economic Assets

SG State Government

SIC Standard Industrial Classification

SN Social Network

Tp Peak Energy Wave Period

TRACE Trade Reporting and Compliance Engine

UN United Nations

US United States

VAR Vector Autoregressive

W Water

WDPs Watershed Development Programs

WP Wave Power

WRDS Wharton Research Data Services

(19)

To all who believed in me

A special thanks goes to my father, Peter, who has supported me

through my entire life, who taught me to work hard for the things that I

aspire to achieve and who prepared me for the challenges of life. My

brother, Markus, who has always been a great resource of help and

advice, my grandparents who helped me through tough times after the

death of my mother as well as Sylvia for her aid throughout the years.

Last but not least, I would like to thank my fiancée, Flordelyn, for her

continuous love and support, especially during the final months of

writing this thesis. I am truly thankful to have you in my life.

(20)
(21)

Part I

(22)
(23)

Chapter 1

Introduction

This PhD thesis uses econometric techniques to provide new insights into energy markets and climate change. In particular, econometric models are applied to the forecasting of oil prices, analyzing the relationship between oil prices and the costs of debt for firms in the US oil indus-try, climate vulnerability and adaptation strategies in the context of smallholder farmers in the Indian watersheds, and long-term trends in the Australian wave climate. Figure 1.1 provides an overview of the structure of this thesis and its six research articles.

The following three sections, Section 1.1, 1.2 and 1.3, will give a brief introduction by out-lining the motivations and objectives for this work, summarizing the research questions, and illustrating the contributions made by each of the six research papers. The main work of this thesis is then divided into three parts. Part II will take a closer look at the oil market and fo-cuses on the broad categories of empirical energy economics and finance. Chapter 2 fofo-cuses on forecasting the real price of oil using forecast combinations, while Chapter 3 is dedicated to the North American oil industry and analyzes how the cost of the debt of oil firms is af-fected by oil price fluctuations. Part III addresses climate change vulnerability and adaptation strategies in the context of smallholder farmers in the Indian watersheds. Chapter 4 lays the theoretical framework by developing a climate vulnerability index and 5 focuses on perceived climate change and the factors that are influencing the choice of adaptation strategies. Part IV is dedicated to the impact of climate change on coastal regions. Thereby, Chapter 6 provides a literature review for the project valuation for coastal adaptation projects and Chapter 7 investi-gates recent trends in the wave climate along the eastern and southeastern coasts of Australia. A short conclusion of the main results of this work and an outlook of possible further research opportunities are provided in part V.

(24)

F IG U R E 1 .1 : Str uctur e of the thesis PhD Thesis Chapter 2

Forecasting the Real

Price of Oil

Chapter 3

Oil Price and Cost of

Debt: Evidence from

Loans and Corporate

Bonds Chapter 7 Long-T erm T rends in the Australian W ave Climate Chapter 6 Project V aluation for Coastal Adaptation Chapter 4 Climate V ulnerability in Rainfed Farming Chapter 5 Changing Climate -Changing Livelihood Part II + III Climate Change Part I

The Oil Market

Cllimate V ulnerability and Adaptation Strategies Coastal Adaptation and W ave Power

Econometric and time

series regression models

Real time data

Out-of-sample forecasts

Forecast uncertainty

Loan and bond market

Panel data

Distributed lag model

Within-between

approach

Theoretical framework

Climate V

ulnerability

Index for Rainfed T

ropics W atershed development programs Rainfed farming Climate change perceptions Adaptation strategies Logit model Bootstrapping Literature review

Cost benefit analysis

Cost ef

fective analysis

Cost utility analysis

Multi-criteria analysis W ave Power Jenson-Shannon divergence Laplacian embedding T rend estimation Bootstrapping

(25)

1.1

Developments in the US Oil Market

The energy landscape is currently in a state of transition and radical changes for energy con-sumption, the electricity and transport sector, and the role of fossil fuels are likely to occur over the next decades. However, the oil market has been and will likely continue to be important in the context of global energy consumption. Figure 1.2 depicts the world energy consumption since 2010 in British thermal unit (Btu) and its projections between 2019 and 2050 as estimated by the U.S. Energy Information Administration (EIA) in the International Energy Outlook 2019 (EIA 2019). Three important trends can be inferred from this figure. First, the total world energy consumption is projected to increase by nearly 50% by 2050. As of 2018, petroleum products and its derivatives make up 32% of the global primary energy consumption (Fig 1.3). Second, renewable energy sources, e.g., solar, wind and hydroelectric power, will increase its share from 15% to 28%. This corresponds to a projected increase by 3.1% per year. Third, while the share of natural gas remains at 22%, coal and petroleum products are projected to have de-creased by 5 and 6 percentage points, respectively. However, the share of petroleum and other liquids is projected to be at 27% of the total energy consumption by 2050. Given the world’s dependence on crude oil as an energy source, the main objective of part II of this thesis is to investigate the oil market and its determinants in more detail.

Since the mid-2000s, there has been a major increase in United States (US) crude oil production mainly driven by the so-called “shale oil revolution”. The term “shale oil” refers to the production of crude oil that is based on rock formations with a low permeability. New extraction methods such as horizontal drilling and hydraulic fracturing (the so-called “fracking”) make it possible to exploit and process crude oil that would have been impossible to release through conventional drilling methods (Kilian 2016). Since the 2008 financial crisis, oil production has increased sharply after a long-lasting decline that started in the 1980s (EIA 2017a). As of 2016, the EIA estimates that 48% of the total US oil production can be attributed

FIGURE1.2: Primary energy consumption by energy source

2010 2020 2030 2040 2050

50

100

150

200

250

Renewables Petroleum and other Liquids

Natural Gas Coal

Nuclear Q u ad ri lli on B ri ti sh t h er m al u n it s

(26)

FIGURE1.3: Share of primary energy consumption by energy source 15% 32% 22% 26% 5% 28% 27% 22% 20% 4%

Renewables Petroleum and other Liquids Natural Gas Coal Nuclear

2018

2018

2050

2050

Note: Own representation based on (EIA 2019).

to shale oil production. Connected to this, Domanski et al. (2015) describe an interesting phenomenon in which there has been a contemporaneous increase in debt-driven investments in the oil sector since the shale oil revolution started. This growth in debt was driven by a macroeconomic environment with low interest rates and investors searching for profitable investments after the financial crisis. This has led to a vicious cycle, where increasing oil prices have led to a higher evaluation of commodity assets and thus increased return expectations. This is particularly important in situations such as that of June 2014, where the market price for crude oil started to decline sharply. This has two main implications for the oil industry: it reduces the cash flow generated by an oil field, which leads to a devaluation of the assets backing a company’s debt. In order to finance their investments, a company could either reduce its assets position or increase its oil production. Both Lehn and Zhu’s (2016) and Lips’s (2019) empirical evidence supports the latter because highly leveraged oil producers cut their investment and increase production to generate cash to pay off their debts. This could lead to an even higher downward pressure on oil prices, which in turn suggests that oil price dynam-ics are affected by an oil company’s capital structure. Consequently, the increasing leverage will lead to a higher risk exposure in the non-financial corporate sector, which may ultimately even put the global financial system at risk (Domanski et al. 2015). Given the latest develop-ment in the crude oil market, Chapter 2 and 3 will try to answer two related research questions. Chapter 2 – Forecasting the Real Price of Oil - Time-Variation and Forecast Combination

This chapter examines the question whether it is possible to generate an accurate forecast of the real price of oil. This is particularly important because central banks, international organiza-tions such as the International Monetary Fund or national instituorganiza-tions like the EIA use real-time forecasts to evaluate possible economic fluctuations as well as to improve macroeconomic pol-icy responses (Baumeister and Kilian 2012, 2015).

This work discusses whether it is possible to generate reliable long-term monthly forecasts of the real price of Brent North Sea Crude Oil (Brent), and how these forecasts could be im-proved by using forecast combinations. The focus of the study also lies on using real time data for constructing out-of-sample forecasts. Thus, the collection of the used data and the challenges in constructing a real-time data set are discussed in detail.

Focusing on econometric and time series regression models, this work contributes to the literature of oil price forecasting by providing an overview of existing models and their per-formance. In addition, some of these existing models will be extended by either introducing more suitable real-time measures for the Brent crude oil price or by using various measures for economic activity. These extensions show improvements in the mean squared prediction error

(27)

(MSPE) ratio in the medium and long run, compared to the original version of these models. Additionally, the performance of the individual models through time is investigated by us-ing recursively constructed MSPE ratios. This demonstrates potential weaknesses of the used models as none of them consistently outperforms the no-change forecasts over time as well as across the different horizons. As such, forecasters have to pay great attention to the sample period they are using. This links directly to the idea of using combination approaches to poten-tially improve the forecasts. Despite having higher MSPE ratios in the long run, a combination of four models outperforms the best individual models over the medium run, while the overall time performance is less volatile.

The second main contribution of this work relates to the forecast uncertainty associated with the point forecasts of the investigated models. By recursively estimating the 5th, 25th, 50th, 75th and 95th quantile of the out-of-sample forecast errors, the study also provides infor-mation about the uncertainty of future forecasts. In addition, this rather simplistic approach is compared to a more sophisticated method using quantile regression analysis. Even though it is possible to construct a forecast that is superior to the no-change benchmark, the uncertainty analysis reveals the limitations of these models.

The Appendix discusses the robustness of the newly introduced models and the four-model combination by using the West Texas Intermediate (WTI) as an alternative benchmark. The WTI is a widely accepted benchmark in comparable forecasting studies. Therefore, the results obtained for the Brent crude oil price are robust and can even be improved when using the WTI. Chapter 3 – Oil Price and Cost of Debt: Evidence from Loans and Corporate Bonds

This chapter examines the effects of oil price fluctuations on the cost of debt for US oil firms. Particularly between January 2011 and June 2014, the crude oil price saw a comparatively stable sideways shift, closely followed by a major decline until January 2015. This could have been the result of a supply shock from the fracking boom in the US as well as OPEC’s refusal to reduce supply quantities. On the other hand, there was evidence that the price depreciation was driven by a demand shock based on a slowdown of the world economy (Baumeister and Kilian 2016). In addition, the appreciation of the US dollar could have lowered the demand even further, as dollar-denominated crude oil imports became more expensive (Baffes et al. 2015). It is particularly interesting that the supply increased even further despite the oil price decline in 2014. This seems to be counterintuitive, as one would expect a cut in production similar to the one after the 2008 financial crisis. By investigating firms’ borrowing decisions and creditworthiness, my colleagues Johannes Lips, Karol Kempa and I will address how the US oil industry was affected by oil price fluctuations.

Thus, in this work, we empirically analyze the relationship between oil price and its volatil-ity and the cost of debt in the US oil industry. Although the impacts of oil price shocks on the world economy and macroeconomic aggregates such as real GDP and real consumption ex-penditure have been extensively studied, little is known about the potential impacts on the oil industry itself. However, this is particularity relevant for the capital-intensive oil industry as oil price fluctuations are likely to have a high impact on the default risk of firms and ultimately their refinancing costs. Thus, we contribute to the literature by investigating whether oil prices, in addition to directly affecting oil firms’ revenues, have an impact at the individual firm level, especially with respect to their cost of debt. We discuss the characteristics of corporate bonds and syndicated loans as two potential sources of debt and the implications the oil price has on their evaluation. We collect data on individual syndicated loans, bond issued by US oil firms, and bonds traded on the secondary market and combine it with data from corporate financial statements. This allows us to analyze how firms’ characteristics such as firm size, profitability, or leverage affect the credit spread of loans and bonds.

To answer the research question, we use a comprehensive database that includes 1,504 pub-licly held firms active in the North American (US and Canada) oil market between Q2 2000 to

(28)

Q1 2018. This includes firm-specific financial variables and detailed information about bond and bank loan data. We categorize these companies according to their position in the value chain in upstream, midstream, downstream and support services firms and discuss the differ-ent potdiffer-ential impacts of oil price and oil price volatility might have on them.

Using two different strategies, we estimate the potential impacts of oil price changes on the cost of debt. The first approach focuses on the costs of debt of individual loans or bonds at issuance, and the second approach takes into account secondary market transactions after the issuance. We use a distributed lag model for estimating the credit spread on the loan and bond level at issuance. Additionally, we use a panel data "within-between"approach to estimate the continuous spreads on the secondary market. This approach is advantageous as it deconstructs the combined effect in the random effect models into between- and within-firm effects. Thus, we can differentiate the effects that oil prices have on a firm depending on its position in the oil industry’s supply chain.

Overall, we find that, even after controlling for loan/bond and firm characteristics as well as the macroeconomic environment, the oil price and its volatility have an effect on a firm’s cost of debt. However, the size and direction of the effects is dependent on the firms’ source of refinancing. For syndicated loans and bonds at issuance we find that higher oil prices in-crease the cost of debt. These findings are intuitively plausible in the case of midstream and downstream firms as crude oil is an input for these firms. Thus, higher oil prices mean higher costs for these firms, which results in higher premiums on the price of debt. Moreover, credit spreads of loans decrease with the volatility of the oil price when considering the full samples of firms, while we find no significant effect for bonds at issuance. This is a surprising result as one would expect banks to charge higher prices for debt in such periods.

When accounting for time-invariant firm heterogeneity, within-between effects estimates for bonds traded on the secondary market indicate that a higher oil price leads to lower fi-nancing costs for firms. In addition, between effects reveal significant differences between the costs of debt for firms in the different industry categories. Overall, the bond market seems to consider high price volatility a risk that increases uncertainty and thus reduces the creditwor-thiness of oil firms. As a result, for credit spreads of bonds traded on the secondary market, we find that firms have to pay higher credit spreads if the volatility increases.

(29)

1.2

Climate Change: Vulnerability and Adaptation Strategies

Climate change is a global phenomenon and the greatest challenge that humankind faces in the current era. According to the IPCC (2018), the global climate has changed relative to the pre-industrial period and, moreover, is already affecting ecosystems and livelihoods worldwide. This includes, among other things, an increase in the land surface temperature (twice the global average) and an increase in the frequency and intensity of some climate and weather extremes, such as extreme temperatures, heavy precipitation, droughts, and heat waves. This poses a threat to world food security and terrestrial ecosystems and contributes to desertification and land degradation in many regions (IPCC 2018, 2019). With the Paris Agreement signed by 197 countries in 2015, the international community has committed itself to taking ambitious effort to combat climate change, adapt to its effects, and enhance the support to developing countries to strengthen their ability to deal with the impacts of climate change (United Nations 2015). To that end, the United Nations (UN) adopted 17 Sustainable Development Goals (SDGs) and 169 targets as part of the 2030 Agenda for Sustainable Development, which provides a 15-year plan to achieve the Goals (UN General Assembly 2015). However, the UN recognized that during the implementation of these goals, too little attention has been given to what should be measured and too little investment has been made to strengthening statistical quality and standards (United Nations Department of Economic and Social Affairs 2019). Part III of my thesis will contribute to the literature by giving attention to climate change vulnerability and adaptation strategies in the context of smallholder farmers in the Indian watersheds.

Expectedly, developing countries in the tropics and subtropics will be hit hardest by climate change (Burney et al. 2014; Hertel and Rosch 2010). This has severe implications for food production and security in these regions as a majority of these countries are characterized by smallholder farming (Harvey et al. 2014). These farmers have a low asset base and operate on less than 2 ha of cropland (World Bank 2003). Of the 3 billion people who live in rural areas, over 80% depend on farming as their main source of income (FAO 2013). Because of their limited available resources, this makes them particularity vulnerable to climate-induced shocks (Frank and Penrose-Buckley 2012; Harvey et al. 2014).

Smallholder farming is prevalent in the Asian continent, with India ranking second in this regard as half of its agricultural production occurs under rainfed farming (FAO 2018). These rainfed areas are characterized by rainfall variability, temperature fluctuations and frequent droughts. Thus, the Indian government decided to initiate watershed development programs (WDPs) to promote this area by providing location-specific support, thus reducing vulnerability, enhancing resilience and building adaptive capacities of rainfed farming communities to climate-induced shocks. Part III of this thesis will focus on two research questions related to WDPs and factors affecting adaptation strategies of smallholder farmers in India.

Chapter 4 – Climate Vulnerability in Rainfed Farming: Analysis from Indian Watersheds

This chapter investigates how climate vulnerability of rainfed farming can be measured and compared. With the global indicator framework adopted by the UN General Assembly in 2017, indicators are recognized as the backbone of monitoring progress toward SDGs at the regional, national and global level (Schmidt-Traub et al. 2015). Thus, my colleagues Archana Raghavan Sathyan, Thomas Aenis, Lutz Breuer and I will contribute to the literature by developing a new kind of Climate Vulnerability Index for Rainfed Tropics (CVIRFT) that is specifically designed to analyze and compare the climate vulnerability of watershed areas. This index can be used to investigate the vulnerability of spatial entities to climate changes in a human dimension. The bottom-up approach described in this work is replicable to similar physio-geographic ar-eas of rainfed farming and can be used to evaluate the potential effectiveness of WDPs used to adapt to climate change. We discuss the concepts used in assessing climate vulnerability and

(30)

follow the widely accepted definition of vulnerability: being a function of exposure, sensitivity and adaptive capacity. Based on an extensive literature review, we modify well-established ap-proaches for estimating climate vulnerability and modify them according to the local situation at the watershed level. The index is comprised of the aforementioned three dimensions and is subdivided into 10 major components. In total, the CVIRFT is based on 59 indicators that capture the determinants of vulnerability.

This study is based on three watersheds established by the Indian government in Kerala, with the goal of promoting development in these areas and addressing the problems associated with climate change for future planning and implementation. We apply the newly developed CVIRFT and test it for comparing the vulnerability of the WDPs that are implemented by three different agencies. We use primary data collected from household surveys (n = 215) to construct the CVIRFT. When comparing the three different implementing agencies, we find a strong variation in the exposure dimension, moderate in sensitivity and a negligible difference in adaptive capacity across the watersheds. After analyzing the major components under the dimensions, we suggest focusing on policy orientation toward redesigning the WDPs. Specifically, we suggest an emphasis on economic diversification, livelihood strategies, social networking coupled with stakeholder participation, natural resource management and risk spread through credit and insurance flexibility.

Chapter 5 – Changing Climate - Changing Livelihood: Smallholder’s Perceptions and Adaptation Strategies

This chapter examines how households perceive climate change and what factors mainly influ-ence their selection of adaptation strategies. This work is an extension of the previous works (Raghavan Sathyan et al. 2016, 2018a,b) that focused on establishing a sensitivity analysis of the CVIRFT. The index is used to compare three watersheds in Kerala where WDPs have been im-plemented. Although the index provides an overall picture of the various levels of vulnerabil-ity, it lacks a detailed analysis of the factors that promote adaptation strategies. My colleagues Archana Raghavan Sathyan, Peter Winker, Lutz Breuer and I will fill this gap and investigate how households perceive climate change, how they adopt their behavior in response to per-ceived changes in climate and what influences their choices of adaption measures. This will help to increase the credibility in examining the impacts of climate change on different scales to find key areas for better policy planning.

For our study site, we find that households perceive an increase in both temperature and the unpredictability of the monsoon season, which is in accordance with actual observed weather trends. Thus, we provide a brief overview of the underlying concepts of perceived climate change and how these perceptions affect farmers’ selection and use of adaptation strategies to cope with climate vulnerability. Households generally use various adaption strategies simul-taneously to cope with such perceptions of climate change. In addition, the three watersheds show considerably different variations in using the actual adaptation strategies implemented.

Based on an extensive literature review, we discuss a broad spectrum of factors that could potentially influence the selection of adaptation strategies used by farmers. These factors are then used as explanatory variables in a binary logistic regression model. Our analysis shows there are various factors that significantly correlate with adaptation strategies used by households to cope with climate change.

(31)

1.3

Climate Change: Coastal Adaptation and Wave Power

Part IV of this thesis is dedicated to the impacts of climate change on coastal regions. Anthropogenic climate change has caused global mean sea levels to rise substantially over the last century. As a result, local and regional sea level variations and the occurrence of sea level extremes increase climate change related risks for coastal regions. This is especially problematic for low-elevation coastal regions, as they are are not only the most attractive in which to live (Kron 2013), but are also characterized by significantly higher populations (Balk et al. 2009) and high asset and infrastructure values. Therefore, coastal flood damages are expected to increase significantly during the 21st century (Hinkel et al. 2014) and are projected to exceed $1 trillion a year by 2050 in major coastal cities if current adaptation measures are not upgraded Hallegatte et al. (2013). Thus, Chapters 6 and 7 will focus on two research questions related climate change’s effects on coastal regions.

Chapter 6 – Project Valuation for Coastal Adaptation – A Literature Review

This chapter presents a literature review of various valuation methods that can be used to evaluate coastal adaptation projects. Given the apparent increase in climate change related risks for coastal regions, the use of suitable adaptation strategies has grown in importance over the last decade. My colleagues Chi Truong, Stefan Trück, Supriya Mathew, and I presents a review of various valuation methods and provide a guide for policy makers on how to evaluate and choose between potential strategies for the management of coastal regions. We discuss cost benefit analysis (CBA), cost effective analysis (CEA), cost utility analysis (CUA) and multi-criteria analysis (MCA) as possible valuation methods and provides examples for their practical application. In addition, an overview of studies that have evaluated coastal adaptation projects using cost benefit and optimal timing approaches is provided to highlight methodological issues in this area of research.

Chapter 7 – Long-Term Trends in the Australian Wave Climate

This chapter examines how the wave climate and extreme values for wave power along Aus-tralia’s coast have changed since the 1970s. My colleague, Stefan Trück, and I contribute to the literature by using a comprehensive data set to investigate the wave climate and extreme values along Australia‘s east and southeast coasts at the regional level. We focus on long-term trends in wave power (and wave height) by using 18 wave rider buoys, ranging between 17 and 42 years of effective record years. We investigate and compare the distributions of wave power by using the Jensen-Shannon divergence and Laplacian embedding simultaneously, which allows us to draw conclusions about the temporal and spatial variations of the wave climate between and within different locations. Through this, we find that stations within a close proximity seem to share similar behavior in terms of their wave power distributions. In addition, we find potential decadal changes in the distribution, which we further address using bootstrapping to test for significant differences. We also illustrate an increase in the yearly mean wave power for a number of locations, including Brisbane, Tweed Head, Coffs Harbour, and Port Kembla. More importantly, we also find an alarming increase in the 99th percentile, or the maximum, over the evaluation period for seven out of the 18 locations. This is of particular interest from a risk management perspective and underlines the importance of location-specific adaptation measures.

(32)
(33)

Part II

(34)
(35)

Chapter 2

Forecasting the Real Price of Oil –

Time-Variation and Forecast

Combination

The following chapter is based on the paper:

Title: Forecasting the Real Price of Oil - Time-Variation and Forecast Combination Authors: Christoph FUNK(contribution: 100%)

Status: Published: Energy Economics, 2018, vol. 76, pp. 288-302 Available from: https://doi.org/10.1016/j.eneco.2018.04.016

Earlier versions of this work were presented at the following scientific conferences with review process:

• 5th International Symposium on Environment and Energy Finance Issues (ISEFI), Paris,

France, May 2017.

• 9thCFE-CMStatistics 2016, Seville, Spain, December 2016.

• 22th Young Researchers Workshops of the German Statistical Society, Augsburg, Germany, September 2016.

(36)

Forecasting the Real Price of Oil –

Time-Variation and Forecast Combination

1

Christoph FUNK2

Abstract

This paper sheds light on the questions whether it is possible to generate an accurate forecast of the real price of oil and how it can be improved using forecast combinations. For this reason, the following paper will investigate the out-of-sample performance of seven individual forecasting models. The results show that it is possible to construct better forecasts compared to a no-change benchmark for horizons up to 24 months with gains in the MSPE ratio as high as 25%. In addition, some of the existing models will be extended, e.g. the US inventories model by introducing more suitable real-time measures for the Brent crude oil price and the VAR model of the global oil market by using different measures for the economic activity. Furthermore, the time performance investigated by constructing recursively estimated MSPE ratios discovers potential weaknesses of the used models. Hence, several different combination approaches are tested with the goal of demonstrating that a combination of individual models is beneficial for the forecasting performance. A combination consisting of four models has proven to have a lower MSPE ratio than the best individual models over the medium run and, in addition, to be remarkably stable over time.

Keywords:Oil price, Forecasting, Combinations, Real-time data, Brent

JEL classification:C53, Q43

1 I am thankful to Peter Winker, Jana Brandt, Daniel Grabowski, Johannes Lips, Regina Ho and Sebastian Probst for their very helpful comments.

2 Faculty of Economics and Business Studies, Department of Statistics and Econometrics, Justus Liebig University Giessen, Licher Str. 64, 35394 Giessen , Germany.

(37)

2.1

Introduction

Crude oil and its derivatives are the most commonly traded commodities in the world. It is widely accepted that the price of oil is a driving factor of the world economy, which is why a reliable forecast of crude oil is important to evaluate possible economic fluctuations3as well as to improve macroeconomic policy responses (Baumeister and Kilian 2012, 2015).

This paper sheds light on the question whether it is possible to generate an accurate fore-cast of the real4price of oil and how it can be improved using forecast combinations. Although most studies have focused on the U.S. Refiners’ Acquisition Cost (RAC) and the WTI as a price benchmark so far, recent research suggests that Brent is the new benchmark that cen-tral bankers monitor and predict. This is mainly due to the fact that the WTI has increasingly suffered from structural instability and thus reflects US rather than global oil market dynamics since 2010/2011 (Baumeister and Kilian 2015; Manescu and van Robays 2016). For this reason, this paper will concentrate on forecasting the Brent oil price using real-time data to generate the forecasts.

This work will investigate the out-of-sample performance of seven individual forecasting models. Some of the existing models will be extended by either introducing more suitable real-time measures for the Brent crude oil price or by using various measures for the economic activity. These extensions show improvements in the MSPE ratio in the medium and long run as compared to the classic versions. However, only the average forecast performance over the complete evaluation period has been addressed with this static indicator. Therefore, the time performance of the investigated models by constructing recursively estimated MSPE ratios discovers potential weaknesses of the used models. None of the analyzed models consistently outperform the no-change forecast over time as well as across the different horizons. Given the difference in the model performance over time, there should at least theoretically exist a model combination that might increase the accuracy of the forecast (Timmermann 2006). Hence, several different combination approaches will be tested with the goal of demonstrating that a combination of individual models is beneficial for the forecasting performance. The results show that it is possible to construct better forecasts as compared to a no-change benchmark for horizons up to 24 months with gains in the MSPE ratio as high as 25%.

Moreover, the used models will be evaluated by means of measuring their forecast uncer-tainty. This is of particular interest, as it allows the forecasters to compare their models based on the uncertainty associated with the obtained point forecasts. Here, the forecast intervals will be computed using a quantile regression approach and quantiles which both are based on previous out-of-sample forecast errors.

The remainder of this paper is organized as follows. Section 2.2 gives a review of recent research and the selected forecasting models considered. Section 2.3 provides a comprehensive description of the used data and the data preparation. Section 2.4 evaluates the forecasting performance of the selected models. Section 2.5 explains the use of forecast combinations and the different approaches that are used to construct them. Section 2.6 extends the analysis of the models by means of an uncertainty analysis. The concluding remarks are in section 2.7.

3 There exists broad evidence in the literature that changes in the price of oil can be used to predict US real GDP growth (see, e.g. Alquist and Kilian (2010), Bachmeier et al. (2008), Hamilton (2011), Kilian and Vigfusson (2011), and Ravazzolo and Rothman (2013) and references therein).

4 Although most of the changes in the nominal price of oil are incorporated in variations of the real price, it is the inflation component that matters when the forecast horizon increases. Besides, when it comes to economic modeling, it is the real price that is the most relevant in that context (Alquist et al. 2013).

(38)

2.2

Selected Forecasting Models

The purpose of this chapter is to present an overview of different models used in this work to predict crude oil prices. While there have been numerous studies5that deal with forecasting the price of oil in general, this study will focus on econometric methods and their predictive power based on time series regression and alternative methods based on oil futures prices, crude oil inventories, non-oil industrial raw materials and oil sensitive stocks.

2.2.1 No-change Benchmark

The random walk without drift is the conventional benchmark in the literature on forecasting asset prices (Alquist et al. 2013). This refers to the idea, that the best forecast for tomorrow is the price of today. Therefore, an h step ahead forecast is set equal to the oil price today:

Ytf+h|t=Yt, (1)

where Ytf+h|tdenotes the h-period forecast of the oil price and Yt refers to the actual real price

of oil. The performance of other models is typically measured relative to this benchmark since a better forecasting performance indicates that the oil price can be predicted at least to some extent. This model will be referred to as the ‘RW’.

2.2.2 AR and ARMA

Autoregressive (AR) and autoregressive moving average (ARMA) processes are two other widely used methods for generating forecasts from time series. Baumeister and Kilian (2012) and Alquist et al. (2013) provide evidence that real oil price forecasts constructed from these models can be more accurate than a no-change benchmark for horizons up to one year eval-uated based on the MSPE and their directional accuracy. Moreover, their results are robust whether the WTI or the RAC is used as the real price of oil in question. As the performance of these two models are very similar, only the results for an ARMA(1,1) model based on the log oil price will be examined in more detail. Additional information about the forecasting performance of the AR model can be found in the supplementary material section.

2.2.3 Futures

Another widely used method is to generate forecasts based on futures prices. This follows the idea that futures contracts might include market expectations about the shift in oil prices. This approach has the advantage that it is relatively simple to generate and easy to communicate to the public (Manescu and van Robays 2016). One can easily generate an h-step forecast for the nominal price of oil by using a futures contract with maturity h. There are several different vari-ations and modificvari-ations discussed in the literature.6Following the suggestion by Baumeister

and Kilian (2012, 2014), the forecast for the real price of oil will be generated by subtracting the expected inflation:

Ytf+h|t=Yt(1+ fth−st−E(πht)), (2)

where Yt is the current level of the real price of oil, fth is the log of current oil futures price

with maturity h, stis the corresponding log of the spot price and E(πht)the expected inflation

rate over the next h periods. This was done by using the monthly average growth rate of the US Consumer Price Index (CPI) up to time t multiplied by h. This approach could be refined

5 The interested reader may find a comprehensive overview of various forecasting techniques in Frey et al. (2009), Bashiri Behmiri and Pires Manso (2013) and Gabralla and Abraham (2013).

(39)

further, but as argued by Baumeister and Kilian (2012, 2014), the deviations are negligible given that the variations of the nominal oil price typically dominate the magnitude of the inflation rate over the horizon of interest.

Although there exists evidence that futures prices are not particularly accurate when it comes to forecasting the oil price (e.g. Alquist and Kilian (2010), Baumeister and Kilian (2012), and Reeve and Vigfusson (2011)), it is still a widely used baseline forecast of central banks (Baumeister and Kilian 2012). Furthermore, when using the Brent oil price as a benchmark, the results of Manescu and van Robays (2016) show slight improvements between 3 and 8 quar-ters ahead forecasts for the time period between 1995Q1 and 2015Q2. However, the results of Baumeister and Kilian (2014) indicate no improvement for up to 4 quarter ahead forecasts for the evaluation period between 1992.01 and 2012.09. This model will be referred to as the ‘Futures’ model.

2.2.4 Forecasts based on Crude Oil Inventories

The fourth class of models considered in this study is based on the theory developed in Alquist and Kilian (2010). A shift in expectations about the real price of oil, ceteris paribus, is reflected in changes in expectations of crude oil inventories. This model has been discussed in Baumeister et al. (2015, 2014) and can be represented as:

Ytf+h|t =Yt(1+βb∆invht), (3)

where ∆invht reflects the percentage change in US crude oil inventories over the preceding h months and bβ is a regression parameter that is obtained by regressing the cumulative

per-centage changes in the real price of oil on the lagged cumulative perper-centage change in US inventories without intercept. Especially, using forecasting horizons of more than 14 months, this approach has been useful in a forecasting combination as it lowers the MSPE as shown in Baumeister and Kilian (2014). Additionally, Baumeister et al. (2015) find MSPE reduction up to 30% as compared to a no-change benchmark between 1992.01 and 2012.09 for monthly data using a MIDAS approach. Results for this classic approach can be found in the supplementary material section. However, this work proposes a new variation of this model by using Organ-isation for Economic Co-operation and Development (OECD) crude oil inventories as there is reason to believe that the US inventories might not be representative for the Brent crude oil price. While there do not exist monthly crude oil inventories data for this model directly, it has been modeled by multiplying US crude oil inventories with the ratio of OECD petroleum stocks over US petroleum stocks. This new approach outperforms the classic model that uses the US inventories. This model will be referred to as the ‘Inventories OECD’ model.

2.2.5 VAR Model of the Global Oil Market

The fifth class of models considered is the Vector Autoregressive (VAR) model of the global oil markets which has been extensively studied over the past several years. According to Baumeis-ter and Kilian (2014), this can be illustrated as a reduced form representation:

B(L)Yt =c+et, (4)

where Yt is a vector of four variables which include the percentage change in global crude oil

production, a measure of global real activity, the log of RAC for crude oil imports deflated by the log of the US CPI, and the change in global crude oil inventories. Furthermore, B(L)denotes the autoregressive lag order polynomial, c a vector of constants, and et represents a vector of

white noise error terms. Only a few works have examined the forecasting ability of this VAR approach for Brent. Baumeister and Kilian (2014) have done this indirectly by modeling the

(40)

spread between Brent and RAC as a random walk and Manescu and van Robays (2016) have modeled the forecasts of Brent directly. However, both methods offer only relatively small or no gains in terms of forecast accuracy as compared to a no-change forecast using quarterly projections.

There exist two different measures for the global real activity that are discussed in the liter-ature so far. Commonly, the "Shipping Index"developed by Kilian (2009), which is constructed from data on global dry cargo ocean shipping freight rates, is used for this purpose. Alterna-tively, one could use the monthly index of industrial production for the OECD and six major non-OECD countries to measure the economic activity. Results for both models can be found in the supplementary material section.

However, two variations of the VAR model will be introduced in the forecasting literature of oil prices by using alternative ways of measuring the global economic activity. Ravazzolo and Vespignani (2015) have shown that the world steel production has a higher predictive accuracy of the world GDP as compared to the OECD industrial production and the Kilian shipping index. Therefore, it seems plausible to use this indicator in order to forecast the real price of Brent crude oil as well. In this case, the log difference of the crude oil price and of the steel production are used in this specification as this outperforms the model using the log-levels. This model will be referred to as the ‘VAR Steel’ model. Additionally, the composite leading indicator of the OECD and six major non-OECD countries will be used to measure the economic activity. This indicator is available on a monthly basis and is designed to detect turning points in business cycles with a lead time of 6 to 9 months. Although it is not a common proxy for the world GDP itself, it might give an advantage for forecasting due to its leading co-movement to the reference series.7 Here, using the log-levels of the oil price and the composite-leading indicator outperforms the specification using log-differences. This model will be referred to as the ‘VAR CLI’ model.

2.2.6 Non-Oil Industrial Raw Materials

An additional way of constructing a real-time forecast would be to use the non-oil industrial raw materials index as discussed in Baumeister and Kilian (2012):

Ytf+h|t =Yt(1+πh,irmt −E(πtn)), (5)

where πth,irm is the percentage change of the Commodity Research Bureau (CRB) index of the spot price of industrial raw materials other than oil over the preceding h months and E(πtn)

is the expected inflation over the same time horizon. Baumeister and Kilian (2012) find MSPE reductions of up to 25% in the short run for this model. Similarly, improvements have been confirmed by Alquist et al. (2013) for the nominal WTI and by Manescu and van Robays (2016) for the real price of Brent. This model will be referred to as the ‘CRB Index’ model.

2.2.7 Price Movements with Oil Sensitive Stocks

S.-S. Chen (2014) has introduced another model that is easy to implement for forecasting the real price of oil. Using monthly data from 1984.10 to 2012.08 S.-S. Chen (2014) shows that using oil sensitive stock price indices can improve the forecasts of nominal and real crude oil prices at short horizons. In particular, the New York Stock Exchange (NYSE) Arca AMEX oil index, which is a price-weighted index of the leading international oil and natural gas companies, can be used to improve the MSPE ratio up to 21% for forecasting the real WTI and real Brent crude oil price. The suggestions made in this paper will be adopted and the model will be referred to as the ‘Oil Stock’ model.

7 In April 2012, the OECD replaced the index of industrial production covering all industry sectors excluding construction with the GDP as its new reference series.

Referenzen

ÄHNLICHE DOKUMENTE

On a long-term basis four regions of Germany will be observed and explored in terms of how climate change and land use change will regionally affect the water cycle,

WE ASK FOR AN INCLUSIVE AND AMBITIOUS CLIMATE REGIME: Initiatives at the national (like Resilient Communities for America or CB27 in Brazil), regional (like the European Covenant of

- the pledge to create an enabling environment that is conducive to making public and private investments consistent with the goals of the Paris Agreement, with an explicit

In order to analyse risks caused by today’s extreme weather events as well as long term effects by global climate change the following methodology was defined..

Within a two-stage non-cooperative game, where industrialized countries decide on funding adaptation in the first stage, and where all countries simultaneously choose

Table S10: Median (lower and upper quartile) change in sugar cane area and production in 2050, expressed as a percentage of the noCC scenario.. In (b), increase (decrease)

USDA-sponsored research continues to support long-term studies to improve understanding of the roles that terrestrial systems play in influencing climate change and the

Sensitivity of runoff t o climatic and other environmental changes is of particular interest t o water resource planners and decision makers because alterations in surface